I have several lines in file, just like below:
/adbc/eee/ddd/baa/ /adbc/fff/ddd/ccc/avfff/ /adbc/ccc/ddd/b/ /adbc/fff/ddd/c/ /adbc/ccc/ddd/bf/ /adbc/ccc/ddd/bc/ The sort algorithm must first get the string before last /, that is:
baa avfff b c bf bc and then sort it by the first character, and then the length of the string, and then alphabetically.
The expected result is
/adbc/fff/ddd/ccc/avfff/ /adbc/ccc/ddd/b/ /adbc/ccc/ddd/bc/ /adbc/ccc/ddd/bf/ /adbc/eee/ddd/baa/ /adbc/fff/ddd/c/ You could use awk in a pre-processing step to add 3 columns all based on the field of the interest, feed to sort, then use cut to discard the extra fields
awk -F'/' -v OFS="/" '{x=substr($(NF-1), 1, 1); print(x, length($(NF-1)), $(NF-1), $0)}' file.txt | sort -k1,1 -k2,2n -k3,3 -t'/' | cut -f4- -d'/' /adbc/fff/ddd/ccc/avfff/ /adbc/ccc/ddd/b/ /adbc/ccc/ddd/bc/ /adbc/ccc/ddd/bf/ /adbc/eee/ddd/baa/ /adbc/fff/ddd/c/ 
cat sortthisfile | while read line do field=$( echo $line | sed -e 's:/$::' -e 's:.*/::' ) firstchar=${field:0:1} fieldlen=${#field} echo "${firstchar},${fieldlen},${field},${line}" done | sort-k1,1 -k2,2n -k3,3 -t, | sed 's:.*,/::' Obviously, sortthisfile is the name of your file.
bash. Surprisingly enough, you sort withsort:-)